Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!ames!sdcsvax!ucbvax!ISUMVS.BITNET!GA.JPH
From: GA.JPH@ISUMVS.BITNET.UUCP
Newsgroups: comp.os.vms
Subject: RE: Buffered I/O count
Message-ID: <8709200642.AA15257@ucbvax.Berkeley.EDU>
Date: Fri, 18-Sep-87 15:56:46 EDT
Article-I.D.: ucbvax.8709200642.AA15257
Posted: Fri Sep 18 15:56:46 1987
Date-Received: Sun, 20-Sep-87 19:37:30 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 27


Todd Warnock <WARNOCK@PRISM.CLEMSON.EDU> writes:

> Where does BUFFERED I/O count come from ?  Does DMA affect it ?  What
> determines when it gets incremented ?  Pointers to manuals would be
> appreciated.  I need to know for accounting purposes.
>

The "Guide to writing a device driver" (or some such title, not having it
front of me at the moment) manual, shows how when writing a device driver
various functions (as in the $QIOW func argument) can be implemented as
buffered or direct I/O.  As a rule fast devices/operations use Direct I/O
slower devices/operations use Buffered.  For example disk devices typically
use DIO while terminal read/writes are BIO (note that [at least on our
system] terminal SETMODE,SENSEMODE, etc. are DIO, lazy coding probably).

For what it's worth, here DIOs are charged for, BIOs are free.

John Hascall
Iowa State University Computation Center
GAJPH@ISUMVS.BITNET

[The opinions expressed are strictly my own and are probably in no way
 related to any official position of ISUCC, Iowa State University, the
 State Board of Regents, the State of Iowa or any sane individual, ...
 although they could be reproduced by an infinite number of monkeys at
 an infinite number of terminals ... ]